Method

Convolution attention point cloud [DKAnet]
[Anonymous Submission]

Submitted on 9 Dec. 2021 02:46 by
[Anonymous Submission]

Running time:0.05 s
Environment:1 core @ 2.0 Ghz (Python)

Method Description:
use Adaptive Convolution and attention on point
cloud 3d object detection
Parameters:
k = 10
Latex Bibtex:

Detailed Results

Object detection and orientation estimation results. Results for object detection are given in terms of average precision (AP) and results for joint object detection and orientation estimation are provided in terms of average orientation similarity (AOS).


Benchmark Easy Moderate Hard
Car (Detection) 95.16 % 93.79 % 89.27 %
Car (Orientation) 95.10 % 93.59 % 88.98 %
Car (3D Detection) 84.57 % 76.70 % 71.54 %
Car (Bird's Eye View) 91.07 % 87.68 % 84.03 %
This table as LaTeX


2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



3D object detection results.
This figure as: png eps pdf txt gnuplot



Bird's eye view results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker